Is 2026 a Good Year to Start a Solo Business? Barely Matters

Every year somebody asks us the same question. Sometimes a reader, sometimes a friend with a corporate job and an idea, increasingly someone who has just been laid off.

Is this a good year to start a solo business?

The honest answer most people give is shaped by what they're selling. Course creators say yes. Career coaches say it depends. We said yes too, in the first version of this article, and backed it with a cost curve nobody had measured.

Then we went looking for actual data. The question turns out to be empirical, the answer is free, and it isn't the one anybody sells you.

The best natural experiment anyone could ask for

The Bureau of Labor Statistics has published survival rates for private-sector establishments by opening year since 1994. It sits in a plain text file called Table 7.

Two rows of it answer the question:

Read those again. Businesses that opened in the year to March 2020 launched straight into a global pandemic, lockdowns, and the sharpest employment collapse in modern record. Five years later, 51.4% of them were still going.

The cohort that opened the year before, into a healthy economy, reached 51.5%.

One tenth of a percentage point. And the pandemic cohort's first-year survival was actually higher — 80.9% against 79.2%.

Across a quarter-century, the spread is about seven points

That isn't a fluke of two adjacent years. Here is five-year survival by opening year across the series:

The best year to have started in the last quarter-century beat the worst by about seven and a half percentage points. BLS's own summary is that startups hitting a recession early generally do worse than those born into recoveries — true, and the effect is single digits.

For comparison, where you start moves things further: one-year survival across the nine census divisions ranges from 71.4% to 84.6%.

So the timing question most of these articles are built on — including ours — is close to noise.

Three things this data doesn't say

It doesn't count you, necessarily. BLS dates a business's birth from its first quarter of positive employment. A one-person business with no payroll may never appear in this series at all. It also counts establishments, not firms — a chain opening a branch is a birth — and a business that reports zero employment temporarily can look like a death and then a rebirth.

It doesn't measure whether the business was any good. Survival and success are different questions, and a business that persists because its owner has no better option still counts as surviving.

And crucially, it says nothing about AI. The 2020 cohort reached its five-year mark in March 2025. ChatGPT launched in November 2022. No cohort has yet completed five years in the generative-AI era — not one. Anyone telling you AI has changed the survival odds is guessing, including anyone telling you it hasn't. That data does not exist yet and cannot for several more years.

That last point is the most honest available answer to "is 2026 different", and it's the one nobody gives.

What the population actually looks like

If the year barely matters, what does the landscape look like for the person asking?

There are about 30.4 million nonemployer businesses in the US — Census counted 30,427,808 in 2023, up from 29,811,495 the year before. They are 78.4% of all US establishments and take in roughly $1.8 trillion, about 6.4% of GDP.

The SBA's Office of Advocacy puts total small businesses at 36.2 million, of which 82.3% have no employees at all. That share has nearly doubled since 1997, from 15.4 million to about 30 million.

Nonemployer businesses grew every single year from 1997 to 2023 except 2008, and grew 4.9% in 2021 and 4.7% in 2022 — the fastest in nearly two decades. Separately, about 16.5 million Americans were self-employed as of June 2026, one in ten of everyone working.

Two honest caveats. Census counts anyone with $1,000 in receipts filing taxes, so this includes a great many people whose "business" is a side income — and the researchers who maintain this data warn it may undercount those filing incorrectly and double-count those filing multiple businesses. It also can't tell you how the work was arranged.

What it pays, which is the part nobody leads with

Here the picture gets sharper, and less flattering.

Census receipts data: about 44.7% of nonemployer firms take under $25,000 a year. Another 22.9% land between $25,000 and $99,900. Just 7.1% clear $100,000.

Payroll data from Gusto on solopreneurs specifically is more encouraging and probably more relevant, because it looks at people who set up properly: 77% profitable in their first year, with owner pay averaging about $41,000 in year one, $76,000 by year two, and over $87,000 by year five. Separate card-data research puts 55% under $50,000 a year and 79% under $100,000.

So: most of these businesses are small. A meaningful minority pay a professional salary. Both things are true and the second one takes about five years.

The "90% fail" number is about a different species

You will see it everywhere and it is not about you.

CB Insights' widely-cited failure research analysed 431 venture-backed companies that shut down since 2023 — 70% ran out of capital, 43% had poor product-market fit. That is a study of venture-backed startups. The 90% figure itself traces to a measure of startups failing to return 10x to investors. A profitable business doing $5 million a year counts as a failure by that definition.

Against that, BLS says roughly half of all establishments reach five years.

And there's a finding that cuts against the received wisdom entirely. Analysis of BLS data published by Chicago Booth found that industries with low barriers to entry — retail, services — had better survival than harder-to-enter sectors like finance and construction. Retail's five-year survival improved from 41% for the 2005 cohort to 56% for the 2011 one.

That argues against the "everything is saturated" panic, ours included. Easy to enter has not historically meant easy to fail.

The barrier that does still bind is licensing, and it changes the shape of the work rather than the odds of survival. If you are entering a licensed trade, the compliance load lands on you personally and on everything your software produces — we took one apart in detail in AI tools for solo real estate agents.

What it actually costs — replacing what we made up

The first version of this article claimed starting cost "$30,000-50,000 in 2010, around $5,000 in 2020, and under $50/month in 2026." We had no source for any of those three numbers. They're gone.

Here is what is actually surveyed. A poll of 700 small-business owners found online-only business owners spent an average of $35,000 in their first year — against $92,500 for mobile businesses and $100,000 for storefronts. More than half said they underestimated what the first year would cost.

On funding, Kauffman Foundation research found at least 83% of entrepreneurs use no bank loan or venture capital at all, and about 65% fund from personal and family savings.

The reconciliation matters more than either figure. The tools really are nearly free — that part of the original claim was right, and we still stand behind a working software stack for under $50 a month, though which AI tools are genuinely cheap for a solo business turns almost entirely on billing basis rather than sticker price. But the tools are not the business. Software is the cheapest line in the budget and the one everyone counts, which is exactly why the average is $35,000 and half of owners are surprised.

"AI changes everything" rests on numbers ranging from 4% to 78%

The original article built two of its four arguments on the assumption that solo founders have already been transformed by AI. That assumption is doing more work than the evidence supports.

The Federal Reserve published three measures of AI adoption in April 2026:

• ~18% of firms have adopted AI in any business function (Census BTOS)
• ~41% of individuals report using generative AI at work
• 78% of the labour force works at a firm that has adopted AI

All three are correct. They count firms, people, and employment-weighted firms respectively.

Underneath that, Census changed its own survey question — from whether businesses use AI to produce goods and services to whether they use it in any business function. The narrow version yields 3.8%. The broad version yields under 20% for firms with four or fewer employees, against 37% for firms of 250+. Ask the US Chamber of Commerce, whose question is whether the owner uses generative AI tools, and you get 58%.

So the honest statement is that measured AI adoption among the very smallest businesses runs anywhere from about 4% to 58%, depending entirely on what you ask. The tools are extraordinary. The claim that they have already reshaped how solo businesses survive is, at present, unmeasured.

The part that got harder, and now has a name

The strongest section of the original article was its least confident one — the observation that supervising AI is not the same as resting, and that "review, correct, approve, review again" produces a fatigue without a name.

It has a name now. A March 2026 Harvard Business Review study, by researchers at Boston Consulting Group and UC Riverside, calls it "AI brain fry" — defined as "mental fatigue resulting from the use or oversight of AI tools beyond an individual's cognitive capacity." Workers reporting it showed higher decision fatigue, more errors, and greater intent to quit. As one analysis of the work put it, AI "has frequently replaced the traditional task with the invisible, high-stakes labor of vetting, fixing, and managing multiple algorithmic outputs."

Two honest notes: the study's specific percentages sit behind a paywall, so we're not quoting them. And it studied employees at large companies — applying it to solo founders is an extrapolation. But the mechanism is now documented rather than merely felt.

And one thing the optimists leave out

The Federal Reserve's Small Business Credit Survey asked 6,525 small employer firms between September and November 2025 how things were going. Current performance held steady — but expectations for revenue and employment growth over the following year fell to their lowest levels since the 2020 survey, and for the second year running slightly more firms reported revenue declines than increases. More than four in ten cited tariff-related costs.

That survey covers firms with employees, not solopreneurs. But it is the largest and most transparent read available on how small businesses feel about the near future, and the answer is: worse than at any point in five years.

So should you start?

Our answer is still broadly yes — but for a completely different reason than we gave before.

Not because 2026 is special. The data says the year is worth a few percentage points at most, and the worst-timed cohort in living memory came out flat. Start because the thing you want to build is something people will pay for, or don't.

The four questions still worth asking:

1. Can you fund the gap? Not "18 months of runway" as a slogan — the real numbers. Owner pay of roughly $41,000 in year one rising to $87,000 by year five is the shape of a working solo business, and 83% of founders cover the gap from personal savings. Budget for a first year that costs more than the software.

2. Are you willing to do distribution? Building has never been cheaper. Being found has never been harder. If posting, writing and showing up publicly is intolerable to you, the odds drop sharply.

3. Does the work benefit from being human-led? If AI can do the whole job, your customer will eventually notice they can do it too.

4. Are you starting toward something or away from something? The research on necessity versus opportunity entrepreneurship consistently finds lower survival for businesses started because there was no alternative — and, more uncomfortably, that necessity founders persist with underperforming ventures because persisting is the only option. If a layoff is what's prompting the question, that's a legitimate reason to start. It's also a reason to be more honest than usual about whether the thing you're building has a buyer.

For what a solo stack really costs, we did the arithmetic in the real cost of a solo AI stack and looked at twelve tools and twelve different meters. And our own account of publishing 30 articles in 90 days is the least varnished thing on this site.

---

Survival figures from the Bureau of Labor Statistics Business Employment Dynamics programme, Table 7 (survival of private sector establishments by opening year, March 1994 – March 2025) and the BLS twentieth-anniversary Spotlight. Business counts from the US Census Bureau Nonemployer Statistics (2023 data) and the SBA Office of Advocacy. Small business sentiment from the Federal Reserve's 2026 Report on Employer Firms (n=6,525). AI adoption figures from the Federal Reserve, the Census Business Trends and Outlook Survey and the US Chamber of Commerce, which measure different things and are reported as such. An earlier version of this article claimed a cost curve, a transaction-volume finding and a time-to-income figure that we had not measured; all are named and removed above. No link in this article pays us — how we make money.